知识卡片

普通读书笔记卡 · 8703.a · 生成模型:GPT-6

句子切分需要领域例外

理解等级:中级

缩写、小数、编号和特殊标点会使按句号切分失败。Punkt等可利用目标域统计和例外规则,但训练后仍须检查典型错误;句子边界也不一定就是最合适的证据边界。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Punkt tokenizer is a tried and tested tool for tokenizing text into sentences. Note that sentence tokenization is not a trivial task, especially if you have domain-specific text. Naive splitting on end marks (periods, question marks, and abbreviations) can only get you so far; abbreviations play spoilsport. You can train the Punkt tokenizer unsupervised over a large body of your target text to ensure it

    参考来源

    - 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html) - 结论依据:摘录中关于句子切分需要领域例外的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Punkt tokenizer is a tried and tested tool for tokenizing text into sentences. Note that sentence tokenization is not a trivial task, especially if you have domain-specific text. Naive splitting on end marks (periods, question marks, and abbreviations) can only get you so far; abbreviations play spoilsport. You can train the Punkt tokenizer unsupervised over a large body of your target text to ensure it